A three-dimensional modeling method and system for scenic spots oriented to tourism management

Abstract

The present invention relates to the field of image data processing technology, and more specifically to a method and system for three-dimensional modeling of scenic spots for tourism management. The method comprises: obtaining a scenic image of a target scenic spot, determining person likelihood information and a Gaussian pyramid within the scenic image; utilizing the person likelihood information to modify the difference images between adjacent layers of the Gaussian pyramid to obtain a target Gaussian difference pyramid; and determining a multi-view stereo model of the target scenic spot based on the target Gaussian difference pyramid. By constructing a Gaussian difference pyramid that removes the influence of tourists, the present invention ensures that the feature points selected by SIFT (Simplified Interference Inference) include, as much as possible, only key points of the scenic spot, excluding key points that may be disrupted by tourists. This improves the accuracy and reliability of the three-dimensional modeling of the scenic spot and avoids deformities, omissions, and distortion during three-dimensional reconstruction.

Description

Technical Field

[0001] The present invention relates to the technical field of image data processing, and in particular to a scenic area three-dimensional modeling method and system oriented to tourism management. Background Art

[0002] 3D models provide intuitive and effective tools for the planning, management, and maintenance of scenic areas. They not only help managers simulate future development and assess the impact of new facilities, but also support the efficient management and allocation of resources. Therefore, 3D modeling of scenic areas is crucial for tourism management. Among various modeling methods, multi-view stereo modeling requires less equipment than laser scanning and other sensor technologies, and with the gradual maturity of data processing technology, its overall cost is more economical.

[0003] Multi-view 3D modeling of scenic spots for tourism management typically requires taking multiple photos of the area using a high-resolution camera or drone. The SIFT (Scale-Invariant Feature Transform) algorithm is then used to extract key points (feature points) from each photo. This is followed by feature matching, camera pose estimation, and depth map generation. Using this multi-view image information, a precise 3D model of the scenic area is reconstructed.

[0004] With existing technologies, the natural environment of scenic spots changes with the seasons, potentially affecting modeling accuracy. Therefore, regular updates to the 3D model are necessary to ensure it reflects the latest scenic environment. However, constantly closing a scenic spot for data collection is unrealistic, especially when tourist activity is unavoidable. Therefore, completely eliminating the influence of tourists when modeling a scenic spot is extremely difficult.

[0005] Visitors' movements within scenic areas are dynamic. If a visitor appears in an image and is mistakenly identified as a feature point, these "false" feature points may be inaccurately matched with feature points from other viewpoints, leading to errors in the matching process. In multi-view stereo modeling, feature point matching errors can affect subsequent geometric calculations. If a visitor is incorrectly matched as a feature point, the error will propagate throughout the entire 3D reconstruction process, ultimately resulting in distorted, missing, or inaccurate 3D models. Summary of the Invention

[0006] In order to solve the technical problem that 3D modeling of scenic spots is easily distorted when affected by people, the present invention aims to provide a 3D modeling method and system for scenic spots for tourism management. The technical solutions adopted are as follows:

[0007] The present invention provides a scenic area three-dimensional modeling method for tourism management, the method comprising:

[0008] Obtaining a scenic area image of the target scenic area, and determining the possible information and Gaussian pyramid of people in the scenic area image;

[0009] Using the person's likelihood information, the difference images between adjacent layers of the Gaussian pyramid are corrected to obtain the target Gaussian difference pyramid.

[0010] Based on the target Gaussian difference pyramid, determine the multi-view stereo modeling of the target scenic area;

[0011] Among them, character possibility information is used to evaluate the impact of character information in the target scenic area on scenic area modeling.

[0012] Furthermore, the person possibility information includes person grayscale possibility; and determining the person possibility information in the scenic area image includes:

[0013] Obtaining a target grayscale value of a target pixel position and a reference grayscale value of a reference pixel position in a scenic area image;

[0014] Determine the grayscale similarity between the target pixel position and the reference pixel position using the target grayscale value and the reference grayscale value;

[0015] The grayscale likelihood of the person in the scenic area image is calculated using the grayscale similarity and the reference image ordinate of the reference pixel position;

[0016] Among them, the person grayscale possibility represents the possibility that the target grayscale value belongs to the person grayscale.

[0017] Furthermore, the person possibility information includes person location possibility; and determining the person possibility information in the scenic area image includes:

[0018] The probability of the person's position in the scenic area image is calculated using the person's grayscale probability and the target image's vertical coordinate of the target pixel position;

[0019] Among them, the person position possibility represents the possibility that the target pixel position belongs to the position of the person.

[0020] Furthermore, the person possibility information includes a single person possibility; and determining the person possibility information in the scenic area image includes:

[0021] determining a target pixel gradient at the target pixel position and a positional proximity between the target pixel position and a reference pixel position;

[0022] The probability of a person's position, target pixel gradient and position proximity are used to calculate the probability of a person in the scenic area image.

[0023] Among them, the single possibility of a person represents the possibility that the pixel gradient change at the target pixel position belongs to the pixel gradient change of the person.

[0024] Furthermore, the scenic area image includes multiple frames of images; the person possibility information includes person comparison possibility; and determining the person possibility information in the scenic area image includes:

[0025] Determine the grayscale difference between any two frames of scenic area images at the same target pixel position;

[0026] The contrast probability of people in the scenic area image is calculated by using the single probability of people and grayscale difference;

[0027] The person contrast possibility represents the possibility that the grayscale change at the target pixel position belongs to the grayscale change of the person.

[0028] Furthermore, the method of using the person possibility information to correct the difference images between adjacent layers of the Gaussian pyramid to obtain a target Gaussian difference pyramid includes:

[0029] Using the character likelihood information, determine the difference coefficient matrix of the Gaussian pyramid;

[0030] The difference coefficient matrix is used to correct the difference images between adjacent layers of the Gaussian pyramid to obtain the target Gaussian difference pyramid.

[0031] Furthermore, the determining of the differential coefficient matrix of the Gaussian pyramid using the character likelihood information includes:

[0032] Using the person possibility information, determine the differential coefficient corresponding to the target pixel position;

[0033] Based on the differential coefficients of each pixel position in the scenic area image, the differential coefficient matrix of the Gaussian pyramid is determined;

[0034] Among them, the character possibility information is negatively correlated with the differential coefficient; the differential coefficient matrix is used to reduce the impact of character information in scenic area images on scenic area modeling.

[0035] Furthermore, the method of using the difference coefficient matrix to correct the difference images between adjacent layers of the Gaussian pyramid to obtain a target Gaussian difference pyramid includes:

[0036] The difference coefficient matrix is used to perform convolution operation on the difference images between adjacent layers of the Gaussian pyramid to calculate the target Gaussian difference pyramid.

[0037] Furthermore, the scenic area image includes a multi-view image; and determining the multi-view stereo modeling of the target scenic area based on the target Gaussian difference pyramid includes:

[0038] Based on the target Gaussian difference pyramid, determine the key points of each perspective image;

[0039] Key points are used for feature matching and three-dimensional reconstruction to obtain a multi-view stereo model of the target scenic area.

[0040] The present invention further provides a scenic area three-dimensional modeling system for tourism management, the system being used to implement the scenic area three-dimensional modeling method for tourism management as described in any one of the above items; the system comprising:

[0041] An image analysis module is used to obtain an image of a target scenic spot and determine the possible information and Gaussian pyramid of people in the image;

[0042] The noise removal module is used to use the person possibility information to correct the difference images between adjacent layers of the Gaussian pyramid to obtain the target Gaussian difference pyramid;

[0043] The model construction module is used to determine the multi-view stereo modeling of the target scenic area based on the target Gaussian difference pyramid.

[0044] The present invention has the following beneficial effects:

[0045] Compared with the existing technology, in which tourists may be mistakenly selected as key points (feature points) of the scenic spot, resulting in errors propagating to the entire 3D reconstruction process, thus causing the final 3D model to be deformed, missing or distorted, the present invention starts from the Gaussian difference pyramid required for SIFT feature extraction of key points, and constructs a Gaussian difference pyramid that removes the influence of tourists, so that the feature points selected by SIFT include only the key points of the scenic spot as much as possible, and do not include key points caused by interference by tourists, thereby improving the accuracy and reliability of 3D modeling of the scenic spot and avoiding deformities, missingness, distortion and the like during 3D reconstruction. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present invention or the prior art, the following is a brief introduction to the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0047] Figure 1 A flowchart of the steps of a three-dimensional modeling method for scenic spots for tourism management provided by one embodiment of the present invention;

[0048] Figure 2 A detailed flow chart of step S1 in a method for three-dimensional modeling of a scenic spot for tourism management provided by one embodiment of the present invention;

[0049] Figure 3 A detailed flow chart of step S1 in a scenic area 3D modeling method for tourism management provided by another embodiment of the present invention;

[0050] Figure 4 A detailed flow chart of step S2 in a method for three-dimensional modeling of a scenic area for tourism management provided by one embodiment of the present invention;

[0051] Figure 5 This is a schematic diagram of the hardware operating environment of a scenic area 3D modeling device for tourism management according to an embodiment of the present invention;

[0052] Figure 6 This is a schematic diagram of the framework structure of a scenic area 3D modeling system for tourism management according to an embodiment of the present invention;

[0053] Figure 7 Schematic diagram of constructing a Gaussian difference pyramid by calculating the differences between adjacent layers of the Gaussian pyramid involved in an embodiment of the present invention. DETAILED DESCRIPTION

[0054] To further illustrate the technical means and effectiveness of the present invention in achieving its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, describes in detail the specific implementation, structure, features, and effectiveness of a three-dimensional modeling method for scenic spots for tourism management proposed in accordance with the present invention. In the following description, references to "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.

[0055] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0056] The following describes in detail a specific solution of a scenic area three-dimensional modeling method for tourism management provided by the present invention in conjunction with the accompanying drawings.

[0057] Example 1:

[0058] For the three-dimensional modeling method of scenic spots for tourism management provided by the present invention, please refer to Figure 1 , which shows a flowchart of the steps of a scenic area three-dimensional modeling method for tourism management provided by an embodiment of the present invention.

[0059] The method comprises:

[0060] Step S1, obtaining a scenic area image of a target scenic area, and determining the person possibility information and Gaussian pyramid in the scenic area image;

[0061] In this embodiment, the scenic area image can be a single image (frame), multiple images of the scenic area, or a panoramic image. To obtain a panoramic image of the target scenic area, a drone can be used to capture different perspectives of the scenic area from the air. At each perspective, multiple frames of images, such as 10 frames, can be captured continuously.

[0062] For any frame of scenic area image, such as the first frame, a Gaussian pyramid can be generated using the SIFT algorithm. A Gaussian pyramid can contain multiple groups of images, each group (Octave) containing multiple images with different blur levels (e.g., 5 layers).

[0063] The person possibility information here refers to the possibility that part of the information in the scenic area image is person information (travelers in the scenic area scene), specifically the possibility that each pixel in the scenic area image is a person pixel. This possibility is used to evaluate the impact of the person information in the target scenic area on the scenic area modeling. The greater the possibility, the greater the interference effect on the scenic area modeling, and vice versa. Therefore, it is necessary to reduce the interference effect of person information on the scenic area modeling.

[0064] Regarding how to determine the likelihood of people in scenic area images, artificial intelligence can be used to determine the person information and corresponding person likelihood information using a neural network model with image recognition capabilities. Alternatively, image feature extraction and object recognition algorithms in the field of computer vision can be used to determine the person likelihood information, such as the SIFT algorithm. This embodiment and the following embodiments primarily utilize the SIFT algorithm and the establishment of a Gaussian pyramid to achieve the determination and recognition of person likelihood information.

[0065] In one embodiment, specifically, the character possibility information includes character grayscale possibility; the character grayscale possibility represents the possibility that the target grayscale value belongs to the character grayscale.

[0066] Please refer to Figure 2 , the step S1 comprises:

[0067] Step S11, obtaining a target grayscale value of a target pixel position and a reference grayscale value of a reference pixel position in the scenic area image;

[0068] Step S12, determining the grayscale similarity between the target pixel position and the reference pixel position using the target grayscale value and the reference grayscale value;

[0069] Step S13 , using the grayscale similarity and the reference image ordinate of the reference pixel position, calculate the grayscale likelihood of the person in the scenic area image.

[0070] When constructing a 3D model of a scenic spot, long-distance shooting helps capture the entire scene or building, while close-up shooting may only capture a local area. Furthermore, when the camera is too close to the scene, especially when using a wide-angle lens, image distortion may occur, resulting in distortion of the object's shape and proportions. Therefore, it is best to maintain a longer distance when shooting.

[0071] When photographing from a distance and at a positive angle, passengers tend to cluster in the lower portion of the image, defined as the region with larger y-values along the y-axis of the image coordinate system. Furthermore, since the colors of the scenery and passengers can differ—for example, scenery (buildings) are generally monotonous and stable, while passengers' clothing can vary in color—we can first determine the likelihood that the grayscale value at each pixel belongs to a passenger. Specifically, for any pixel in the image, if the grayscale value of another pixel is similar, and the y-value of that pixel is larger, then the grayscale at that pixel is more likely to belong to a passenger.

[0072] Construct the following formula to represent the position of any target pixel in the scenic area image: The possibility that the grayscale value above belongs to the passenger grayscale, that is, the character grayscale possibility:

[0073] ;

[0074] Formula explanation: Where, Indicates the target pixel position in the scenic area image The possibility that the grayscale value above belongs to the grayscale of the tourist, that is, the grayscale possibility of the person; m represents the x-axis range in the scenic area image, n represents the y-axis range in the scenic area image, and exp represents the exponential function with e as the base. Indicates the target pixel position in the scenic area image The target grayscale value on Indicates the reference pixel position in the scenic area image The reference gray value on Indicates the target pixel position in the scenic area image With reference pixel position The grayscale similarity of Indicates the reference pixel position Similarly, we can determine the possibility that the grayscale value at each pixel position in the scenic area image belongs to the tourist grayscale.

[0075] Based on the above embodiments, further, in one embodiment, the person possibility information includes person position possibility, and the person position possibility represents the possibility that the target pixel position belongs to the position where the person is located.

[0076] The step S1 includes:

[0077] The probability of the person's position in the scenic area image is calculated using the person's grayscale probability and the target image's ordinate of the target pixel position.

[0078] The above example has determined the likelihood that the grayscale value at any pixel location belongs to a passenger. However, the same grayscale value may be distributed at different locations in the image, and the likelihood is not the same at all locations. Only when the grayscale value at a pixel location has a higher grayscale likelihood and the y value at that pixel location is also large, can it be more strongly indicated that the pixel location is more likely to belong to a passenger.

[0079] Construct the following formula to represent the target pixel position in the scenic area image Possible character positions on:

[0080] ;

[0081] Formula explanation: Indicates the pixel position in the scenic area image The probability of the character position on the target pixel; v represents the target pixel position The y value (the vertical coordinate of the target image), Indicates the grayscale possibility of a person.

[0082] Similarly, the probability of the passenger's position at each pixel position in the scenic area image can be determined.

[0083] Based on the above embodiments, further, in one embodiment, the person possibility information includes a single person possibility, and the single person possibility represents the possibility that the pixel gradient change at the target pixel position belongs to the pixel gradient change of the person.

[0084] The step S1 includes:

[0085] determining a target pixel gradient at the target pixel position and a positional proximity between the target pixel position and a reference pixel position;

[0086] The single probability of a person in the scenic area image is calculated using the person's position probability, target pixel gradient and position proximity.

[0087] The person position probability obtained in the above embodiment may not fully represent the possibility of a passenger. For example, areas such as land and lawn may also lead to a higher person position probability at the corresponding pixel position.

[0088] However, because land and lawns have a high degree of sprawl (less prone to pixel gradient changes), while tourists' grayscale values have a low degree of sprawl (more prone to pixel gradient changes), even though these areas may have similar grayscale probability locations to tourists, meaning they have a high probability of being human, their grayscale value variation patterns differ from those of tourists. Specifically, for each pixel in the scenic area image, if the pixel gradient variation based on the human location probability is significant, then the probability of that pixel belonging to a tourist is high; otherwise, it is low. By combining the gradient information of grayscale changes, the influence of factors such as lawn and land can be removed, thereby improving the accuracy of passenger identification.

[0089] Construct the following formula to represent the target pixel position in the scenic area image The single possibility of the above characters:

[0090] ;

[0091] ;

[0092] Formula explanation: Indicates the target pixel position in the scenic area image The single possibility of the character on the image, m represents the x-axis range in the scenic area image, n represents the y-axis range in the scenic area image, Indicates the target pixel position in the scenic area image With reference pixel position Position proximity, u represents the target pixel position The x value, v represents the target pixel position The y value of a represents the reference pixel position. The x value of b represents the reference pixel position. The y value of represents a hyperparameter (making the denominator non-zero), Indicates the target pixel position The target pixel gradient, which may specifically be a grayscale gradient; Indicates the possible positions of characters.

[0093] Similarly, the single possibility of a traveler (person) at each pixel position in the scenic area image can be determined.

[0094] Based on the above embodiments, further, in one embodiment, the scenic area image includes multiple frames of images; the person possibility information includes person contrast possibility, and the person contrast possibility represents the possibility that the grayscale change at the target pixel position belongs to the grayscale change of the person.

[0095] Please refer to Figure 3 , the step S1 comprises:

[0096] Step S101, determining the grayscale difference between any two frames of scenic area images at the same target pixel position;

[0097] Step S102 : Calculate the contrast probability of people in the scenic area image using the single probability of the people and the grayscale difference.

[0098] Since scenery is relatively static and tourists are relatively dynamic, for each pixel position in the scenic area image, if it belongs to a tourist, then based on the single possibility of a tourist in a single image, if there is a significant grayscale change at this pixel position in the next N frames, it can be further explained that the contrast possibility of this pixel position belonging to a tourist is high.

[0099] Construct the following formula to represent the target pixel position in the scenic area image Possibility of comparing characters on:

[0100] ;

[0101] Formula explanation: Indicates the target pixel position in the scenic area image The possibility of character comparison on the image, N represents the number of image frames captured at the current viewing angle. For example, N=10 means that 10 frames of scenic area images are captured at the current viewing angle. Indicates the target pixel position in the t-th frame image The gray value on Indicates the target pixel position in the scenic area image (can be the first frame) The gray value on Represents the grayscale difference between any two frames of scenic area images at the same target pixel position, Indicates the target pixel position in the scenic area image The single possibility of the character above.

[0102] Similarly, the possibility of character comparison at all pixel positions in the scenic area image can be determined.

[0103] Step S2: Using the person likelihood information, correct the difference images between adjacent layers of the Gaussian pyramid to obtain the target Gaussian difference pyramid;

[0104] For details, please refer to Figure 4 , the step S2 comprises:

[0105] Step S21, using the character likelihood information, determining the difference coefficient matrix of the Gaussian pyramid;

[0106] Specifically, the step S21 includes:

[0107] Using the person possibility information, determine the differential coefficient corresponding to the target pixel position;

[0108] Based on the differential coefficients of each pixel position in the scenic area image, the differential coefficient matrix of the Gaussian pyramid is determined;

[0109] Among them, the character possibility information is negatively correlated with the differential coefficient; the differential coefficient matrix is used to reduce the impact of character information in scenic area images on scenic area modeling.

[0110] Taking the first frame of the scenic spot image as an example, when constructing the Gaussian difference pyramid (DOG) of the original group in the first frame of the image, Figure 7 As shown, a differential calculation is performed on all pixel positions in adjacent layers of the original group in the Gaussian pyramid to obtain differential values and generate a differential image. The magnitude of the differential value directly influences the subsequent determination of local extreme points. For the differential value at the passenger's pixel location, a large differential value does not necessarily indicate an extreme point that truly reflects the characteristics of the scenic spot. Therefore, a smaller differential value can be assigned to the passenger's pixel location to reduce its interference with scenic spot feature extraction.

[0111] Because when constructing the Gaussian difference pyramid for the original group, if the passenger at a certain target pixel position has a greater likelihood of contrast or other person likelihood information relative to other pixel positions, then the probability of that target pixel position being a scenic spot is low, and a smaller difference coefficient should be given when calculating the difference image; conversely, if the passenger at a certain target pixel position has a lower likelihood of contrast or other person likelihood information relative to other pixel positions, then the probability of that target pixel position being a scenic spot is high, and a larger difference coefficient should be given to that target pixel position. In this way, the following matrix can be constructed to represent the difference coefficient matrix of the original group of the Gaussian pyramid:

[0112] ;

[0113] Represents the differential coefficient matrix of the original group of the Gaussian pyramid, where each element represents the differential coefficient corresponding to each (target) pixel position. Similarly, the differential coefficient matrix of each group of the Gaussian pyramid can be obtained. f represents the maximum and minimum normalization function, that is, Min-Max Normalization. The character contrast possibility here Can be replaced with a single possibility of character Or character location possibilities Or character grayscale possibility , optimal character comparison possibility , priority > > > In step S22 , the difference coefficient matrix is used to correct the difference images between adjacent layers of the Gaussian pyramid to obtain a target Gaussian difference pyramid.

[0114] Specifically, the step S22 includes:

[0115] The difference coefficient matrix is used to perform convolution operation on the difference images between adjacent layers of the Gaussian pyramid to calculate the target Gaussian difference pyramid.

[0116] Here we take the original group as an example to illustrate. The above embodiment has determined the difference coefficient matrix of the original group Therefore, when performing differentiation on any two adjacent layers in the original group of the Gaussian pyramid, the differential coefficient matrix can be convolved with the differential image to obtain a modified differential image after removing the influence of passengers.

[0117] Based on the above description, we can first construct the following formula to express the corrected difference image between the first and second layers of the original Gaussian pyramid group:

[0118] ;

[0119] Formula explanation: represents the modified difference image between the first and second layers of the original set of Gaussian pyramids, represents the difference coefficient matrix of the original group, represents the convolution operation, represents the first layer image of the original group of Gaussian pyramid, represents the second layer image of the original group of Gaussian pyramid, Represents the initial difference image between the first and second levels of the original set of Gaussian pyramids.

[0120] Similarly, the difference images of all adjacent layers in the original group of the Gaussian pyramid of the first frame image (other frame images can also be selected) can be constructed to obtain the Gaussian difference pyramid corresponding to the original group.

[0121] Because the next set of images in the original Gaussian pyramid is created through downsampling, the pixel positions of the passengers in the original set are also downsampled, so the influence of the passenger pixel positions on the differential coefficients is also eliminated through downsampling. Therefore, the Gaussian difference pyramid corresponding to the next set in the Gaussian pyramid of the first frame (or other frames) can be determined by the following steps:

[0122] First, the difference coefficient matrix of the original group Downsampling can get the next group of differential coefficient matrix .

[0123] Similar to the above steps, the following formula is used to construct the corrected difference image between the first and second layers of the next group of Gaussian pyramids:

[0124] ;

[0125] Formula explanation: Represents the corrected difference image between the first and second layers of the next group of Gaussian pyramids, represents the next set of differential coefficient matrix, represents the convolution operation, Represents the first layer image of the next group of Gaussian pyramids, Represents the second layer image of the next group of Gaussian pyramids, Represents the initial difference image between the first and second layers of the next group.

[0126] Similarly, we can construct the difference images of all adjacent layers in the next group of the Gaussian pyramid of the first frame image, and thus obtain the Gaussian difference pyramid corresponding to this group. By repeating the above operations for all groups of the Gaussian pyramid from bottom to top, we can gradually construct the complete target Gaussian difference pyramid that removes the influence of passengers.

[0127] Step S3, determining a multi-view stereo model of the target scenic area based on the target Gaussian difference pyramid;

[0128] Specifically, step S3 includes:

[0129] Based on the target Gaussian difference pyramid, determine the key points of each perspective image;

[0130] Key points are used for feature matching and three-dimensional reconstruction to obtain a multi-view stereo model of the target scenic area.

[0131] Based on the above embodiment, the target Gaussian difference pyramid for removing the passenger effect in the first frame image has been constructed. Next, the key points in the first frame image are extracted through the remaining steps of the following SIFT algorithm:

[0132] 1. Find local extreme points: In each layer and adjacent layers, detect local extreme points as preliminary key points.

[0133] 2. Accurately position key points: Further improve the position accuracy of key points by fitting methods and eliminating unstable points.

[0134] 3. Assign direction: Assign a rotation-invariant direction to each keypoint based on the local gradient direction.

[0135] 4. Generate descriptors: By calculating the gradient direction histogram of the area around the key point, a descriptor is generated to describe the local features of the area.

[0136] In this way, through the above steps, non-passenger key points with high robustness can be extracted under different scales, different rotation angles, and different lighting conditions.

[0137] Obtain the key points (feature points) of the scenic area that are not tourists from all perspectives. Next, perform multi-perspective stereo modeling:

[0138] 1. Feature matching: The same feature points in images from different perspectives are paired using a feature matching algorithm based on Euclidean distance.

[0139] 2. Camera pose estimation: Using camera intrinsic parameters and feature matching points, triangulation is used to estimate the camera position and orientation for each viewpoint.

[0140] 3. Depth map generation: Based on the camera pose and matched feature points, the distance from each pixel to the camera is calculated to generate a depth map.

[0141] 4. 3D reconstruction: Combine multi-view depth maps to calculate the 3D coordinates of points in the scene and generate a 3D point cloud.

[0142] 5. Surface reconstruction: Using the Poisson Surface Reconstruction method, a complete surface model is constructed through the connection relationship of the point cloud.

[0143] 6. Texture mapping: Projecting the texture information (such as color and details) of the captured image onto the surface of the 3D model.

[0144] 7. 3D model optimization and refinement: Through multiple reconstructions and optimizations, the accuracy and details of the model are improved, which can accurately reflect the real situation of the scenic area and reduce the impact of tourists on the scenic area modeling.

[0145] Compared with the existing technology, in which tourists may be mistakenly selected as key points (feature points) of the scenic spot, resulting in errors propagating to the entire 3D reconstruction process, thus causing the final 3D model to be deformed, missing or distorted, the present invention starts from the Gaussian difference pyramid required for SIFT feature extraction of key points, and constructs a Gaussian difference pyramid that removes the influence of tourists, so that the feature points selected by SIFT include only the key points of the scenic spot as much as possible, and do not include key points caused by interference by tourists, thereby improving the accuracy and reliability of 3D modeling of the scenic spot and avoiding deformities, missingness, distortion and the like during 3D reconstruction.

[0146] Example 2:

[0147] The embodiment of the present invention further provides a scenic spot 3D modeling device for tourism management. The scenic spot 3D modeling device for tourism management can be a data computing and processing device such as a computer, a server, or a combination of multiple devices.

[0148] like Figure 5 As shown, Figure 5It is a structural diagram of the hardware operating environment of the scenic area three-dimensional modeling device for tourism management involved in the embodiment of the present invention.

[0149] like Figure 5 As shown, the scenic area 3D modeling device for tourism management may include: a processor 1001, such as a CPU, a network interface 1004, a user interface 1003, a memory 1005, and a communication bus 1002. Communication bus 1002 is used to enable communication between these components. User interface 1003 may include a display and an input unit, such as a control panel. Optionally, user interface 1003 may also include a standard wired interface or a wireless interface. Network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface). Memory 1005 may be high-speed RAM or non-volatile memory, such as a disk drive. Memory 1005 may also be a storage device independent of processor 1001. Memory 1005, a computer storage medium, may include a scenic area 3D modeling program.

[0150] Those skilled in the art will understand that Figure 5 The hardware structure shown in the figure does not constitute a limitation of the device, and may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently.

[0151] Continue to refer to Figure 5 , Figure 5 The memory 1005 as a computer-readable storage medium may include an operating system, a user interface module, a network communication module, and a scenic area three-dimensional modeling program.

[0152] exist Figure 5 In the embodiment, the network communication module is mainly used to connect to the server and can communicate data with the server; and the processor 1001 can call the scenic area three-dimensional modeling program stored in the memory 1005 and execute the steps in the above embodiments.

[0153] The hardware structure of the above-mentioned scenic area 3D modeling device for tourism management is used to implement various embodiments of the scenic area 3D modeling method for tourism management of the present invention.

[0154] In addition, the present invention also provides a scenic area 3D modeling system for tourism management, please refer to Figure 6 The scenic area 3D modeling system for tourism management includes:

[0155] Image analysis module A10, used to obtain a scenic area image of the target scenic area, and determine the person possibility information and Gaussian pyramid in the scenic area image;

[0156] The noise removal module A20 is used to use the person possibility information to correct the difference images between adjacent layers of the Gaussian pyramid to obtain the target Gaussian difference pyramid;

[0157] The model construction module A30 is used to determine the multi-view stereo modeling of the target scenic area based on the target Gaussian difference pyramid.

[0158] Furthermore, the image analysis module A10 is further configured to:

[0159] Obtaining a target grayscale value of a target pixel position and a reference grayscale value of a reference pixel position in a scenic area image;

[0160] Determine the grayscale similarity between the target pixel position and the reference pixel position using the target grayscale value and the reference grayscale value;

[0161] The grayscale likelihood of the person in the scenic area image is calculated using the grayscale similarity and the reference image ordinate of the reference pixel position;

[0162] Among them, the person grayscale possibility represents the possibility that the target grayscale value belongs to the person grayscale.

[0163] Furthermore, the image analysis module A10 is further configured to:

[0164] The probability of the person's position in the scenic area image is calculated using the person's grayscale probability and the target image's vertical coordinate of the target pixel position;

[0165] Among them, the person position possibility represents the possibility that the target pixel position belongs to the position of the person.

[0166] Furthermore, the image analysis module A10 is further configured to:

[0167] determining a target pixel gradient at the target pixel position and a positional proximity between the target pixel position and a reference pixel position;

[0168] The probability of a person's position, target pixel gradient and position proximity are used to calculate the probability of a person in the scenic area image.

[0169] Among them, the single possibility of a person represents the possibility that the pixel gradient change at the target pixel position belongs to the pixel gradient change of the person.

[0170] Furthermore, the image analysis module A10 is further configured to:

[0171] Determine the grayscale difference between any two frames of scenic area images at the same target pixel position;

[0172] The contrast probability of people in the scenic area image is calculated by using the single probability of people and grayscale difference;

[0173] The person contrast possibility represents the possibility that the grayscale change at the target pixel position belongs to the grayscale change of the person.

[0174] Furthermore, the noise elimination module A20 is further configured to:

[0175] Using the character likelihood information, determine the difference coefficient matrix of the Gaussian pyramid;

[0176] The difference coefficient matrix is used to correct the difference images between adjacent layers of the Gaussian pyramid to obtain the target Gaussian difference pyramid.

[0177] Furthermore, the noise elimination module A20 is further configured to:

[0178] Using the person possibility information, determine the differential coefficient corresponding to the target pixel position;

[0179] Based on the differential coefficients of each pixel position in the scenic area image, the differential coefficient matrix of the Gaussian pyramid is determined;

[0180] Among them, the character possibility information is negatively correlated with the differential coefficient; the differential coefficient matrix is used to reduce the impact of character information in scenic area images on scenic area modeling.

[0181] Furthermore, the noise elimination module A20 is further configured to:

[0182] The difference coefficient matrix is used to perform convolution operation on the difference images between adjacent layers of the Gaussian pyramid to calculate the target Gaussian difference pyramid.

[0183] Furthermore, the model construction module A30 is also used to:

[0184] Based on the target Gaussian difference pyramid, determine the key points of each perspective image;

[0185] Key points are used for feature matching and three-dimensional reconstruction to obtain a multi-view stereo model of the target scenic area.

[0186] The specific implementation of the scenic area 3D modeling system for tourism management of the present invention is basically the same as the embodiments of the scenic area 3D modeling method for tourism management described above, and will not be repeated here.

[0187] The present invention also provides a computer-readable storage medium having a scenic area 3D modeling program stored thereon, wherein when the scenic area 3D modeling program is executed by a processor, the steps of the above-mentioned scenic area 3D modeling method for tourism management are implemented.

[0188] Among them, the method implemented when the scenic area three-dimensional modeling program is executed can refer to the various embodiments of the scenic area three-dimensional modeling method for tourism management of the present invention, and will not be repeated here.

[0189] It should be noted that the order in which the embodiments of the present invention are described above is for illustrative purposes only and does not necessarily represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or sequential order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0190] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on the differences from the reference embodiment.

[0191] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0192] The above description is only a preferred embodiment of the present invention and does not limit the scope of protection of the present invention. All equivalent structural / method transformations made by using the contents of the present invention description and drawings under the inventive concept of the present invention, or direct / indirect application in reference to related technical fields are included in the scope of protection of the present invention.

Claims

1. A three-dimensional modeling method for scenic spots for tourism management, characterized in that: The method comprises: Acquire a scenic area image of a target scenic area, and determine person likelihood information and a Gaussian pyramid in the scenic area image, wherein the person likelihood information includes person grayscale likelihood; Using the person's likelihood information, the difference images between adjacent layers of the Gaussian pyramid are corrected to obtain the target Gaussian difference pyramid. Based on the target Gaussian difference pyramid, determine the multi-view stereo modeling of the target scenic area; Among them, character possibility information is used to evaluate the impact of character information in the target scenic spot on scenic spot modeling; The scenic area image includes a multi-view image; and determining the multi-view stereo modeling of the target scenic area based on the target Gaussian difference pyramid includes: Based on the target Gaussian difference pyramid, determine the key points of each perspective image; Use key points to perform feature matching and 3D reconstruction to obtain a multi-view stereo model of the target scenic area; Determining the possible person information in the scenic area image includes: Obtaining a target grayscale value of a target pixel position and a reference grayscale value of a reference pixel position in a scenic area image; Determine the grayscale similarity between the target pixel position and the reference pixel position using the target grayscale value and the reference grayscale value; The grayscale probability of the person in the scenic area image is calculated using the grayscale similarity and the reference image ordinate of the reference pixel position; Among them, the person grayscale possibility represents the possibility that the target grayscale value belongs to the person grayscale.

2. The scenic area three-dimensional modeling method for tourism management according to claim 1 is characterized in that: The character possibility information includes character location possibility; Determining the possible person information in the scenic area image includes: The probability of the person's position in the scenic area image is calculated using the person's grayscale probability and the target image's vertical coordinate of the target pixel position; Among them, the person position possibility represents the possibility that the target pixel position belongs to the position of the person.

3. The scenic area three-dimensional modeling method for tourism management according to claim 2, characterized in that: The character possibility information includes a single possibility of a character; Determining the possible person information in the scenic area image includes: determining a target pixel gradient at the target pixel position and a positional proximity between the target pixel position and a reference pixel position; The probability of a person's position, target pixel gradient and position proximity are used to calculate the probability of a person in the scenic area image. Among them, the single possibility of a person represents the possibility that the pixel gradient change at the target pixel position belongs to the pixel gradient change of the person.

4. The scenic area three-dimensional modeling method for tourism management according to claim 3 is characterized in that: The scenic area image includes multiple frames of images; the character possibility information includes character comparison possibility; Determining the possible person information in the scenic area image includes: Determine the grayscale difference between any two frames of scenic area images at the same target pixel position; The contrast probability of people in the scenic area image is calculated by using the single probability of people and grayscale difference; The person contrast possibility represents the possibility that the grayscale change at the target pixel position belongs to the grayscale change of the person.

5. The scenic area three-dimensional modeling method for tourism management according to claim 1, characterized in that: The method of using the person possibility information to correct the difference images between adjacent layers of the Gaussian pyramid to obtain a target Gaussian difference pyramid includes: Using the character likelihood information, determine the difference coefficient matrix of the Gaussian pyramid; The difference coefficient matrix is used to correct the difference images between adjacent layers of the Gaussian pyramid to obtain the target Gaussian difference pyramid.

6. The scenic area three-dimensional modeling method for tourism management according to claim 5, characterized in that: The method of determining the differential coefficient matrix of the Gaussian pyramid by using the character likelihood information includes: Using the person possibility information, determine the differential coefficient corresponding to the target pixel position; Based on the differential coefficients of each pixel position in the scenic area image, the differential coefficient matrix of the Gaussian pyramid is determined; Among them, the character possibility information is negatively correlated with the differential coefficient; the differential coefficient matrix is used to reduce the impact of character information in scenic area images on scenic area modeling.

7. The scenic area three-dimensional modeling method for tourism management according to claim 5, characterized in that: The method of using the difference coefficient matrix to correct the difference images between adjacent layers of the Gaussian pyramid to obtain a target Gaussian difference pyramid includes: The difference coefficient matrix is used to perform convolution operation on the difference images between adjacent layers of the Gaussian pyramid to calculate the target Gaussian difference pyramid.

8. A scenic area 3D modeling system for tourism management, characterized by: The system is used to implement the scenic area three-dimensional modeling method for tourism management according to any one of claims 1 to 7; the system includes: An image analysis module is used to obtain an image of a target scenic spot and determine person likelihood information and a Gaussian pyramid in the image, wherein the person likelihood information includes person grayscale likelihood; The noise removal module is used to use the person possibility information to correct the difference images between adjacent layers of the Gaussian pyramid to obtain the target Gaussian difference pyramid; A model construction module is used to determine the multi-view stereo modeling of the target scenic area based on the target Gaussian difference pyramid; The scenic area image includes a multi-view image; and determining the multi-view stereo modeling of the target scenic area based on the target Gaussian difference pyramid includes: Based on the target Gaussian difference pyramid, determine the key points of each perspective image; Use key points to perform feature matching and 3D reconstruction to obtain a multi-view stereo model of the target scenic area; Determining the possible person information in the scenic area image includes: Obtaining a target grayscale value of a target pixel position and a reference grayscale value of a reference pixel position in a scenic area image; Determine the grayscale similarity between the target pixel position and the reference pixel position using the target grayscale value and the reference grayscale value; The grayscale probability of the person in the scenic area image is calculated using the grayscale similarity and the reference image ordinate of the reference pixel position; Among them, the person grayscale possibility represents the possibility that the target grayscale value belongs to the person grayscale.

Images